Methods Of Clutter Suppression And Target Detection For Airborne Radar Space-Time Processing

Airborne radar is usually working in a downward-looking state and the received backscattered echoes often contain strong ground clutter which will be broadened on the range-Doppler plane.Thus,weak targets are often absorbed in the clutter which will decrease the radar detection performance.Space-time adaptive processing(STAP)technology can suppress clutter effectively and improve the radar’s target detection performance.STAP technology requires a large number of independent and identically distributed(i.i.d)training samples to estimate the clutter covariance matrix of the cell under test(CUT).In the actual clutter environment,it’s pretty difficult to obtain enough i.i.d samples,which will lead to the inaccurate estimation of clutter covariance matrix and degradation of clutter suppression inevitably.In the non-uniform clutter environment,by studing the sparse representation technology under the condition of small samples and selecting training samples based on prior knowledge,this paper is mainly improving the refined division method of RangeDoppler cells and improving the convergence rate of sparse Bayesian learning The main content of the paper is summarized as follows:1.The method of weighting training samples after sample selection and covariance matrix estimation is studied.Aiming at the problem that traditional range Doppler unit subdivision method will harm the performance of long-range clutter suppression,the following improvements are made to improve it: Firstly,the number of divisions corresponding to the average area of all range-Doppler units is obtained.Then the division number of the remaining distance Doppler units is determined by the ratio of their area.The backscattering coefficients of each subdivision is obtained according to the clutter statistical characteristics under different surface coverage types.and training samples are selected through the backscattering coefficient distribution.Finnaly,similarity between the training samples and the CUT is used to obtain the corresponding weight of each training sample and the estimated sample covariance matrix is obtained.This method takes the clutter distribution in all rangeDoppler units into account and improves the clutter suppression performance of the STAP algorithm in a non-uniform environment.Experimental results using the real data show that the remote clutter suppression capability of this mehtod is improved by more than 3d B compared with the traditional sample selection methods.2.The clutter suppression method of fast convergent sparse Bayesian learning under the condition of small samples is studied.In order to solve the problem of huge computation which exists in the iterative process of the existing sparse Bayesian learning algorithm.The existing fast-convergent sparse Bayesian learning algorithm is improved from two aspects:the selection of dictionary matrix and the simplified matrix inversion operation.First,using the direct data domain(DDD)algorithm and prior knowledge to determine the clutter ridge position,the proposed method selects the grid points near the clutter ridge to form a dictionary matrix.Then this method introduces a unitary matrix to further reduce the amount of calculation for matrix inversion and speeds up the algorithm Convergence speed.Finally the proposed method can get a more accurate covariance matrix.This method greatly reduces the amount of calculation and increases the speed of calculation within the range where the clutter suppression performance is allowed to decrease.